Code Quality in the Age of Coding Agents
An adaptation of an internal talk I gave during a company wide ai-adoption week.
Welcome to my website!
An adaptation of an internal talk I gave during a company wide ai-adoption week.
Software is highly checkable, but end-to-end verification is hard because deciding what should exist is the real work. A critique of AI verification through the lens of Naur, Deutsch, and Alexander.
Sports betting might seem an unconventional AI benchmark, yet it uniquely tests reasoning under uncertainty, decision-making transparency, and real-world information synthesis.
I've created two new SolidJS applications in the last three months. Both times I had issues with Layouts caused by a failure to consider SolidJS reactivity model. Here is a reference to remind future me why I keep making this mistake.
When building a product company is it better to keep building more features and broadening the product offering, or focus on doing fewer things but being the absolute best at those things.
There is real risk in failing to innovate and build. Growing too slow may be a slow death, but it is a very real failure mode. If you have a lot of technical debt in your product, throwing it away and starting again may be risky, but it may be less risky in the long term than persisting with a broken foundation.
There are a surprising number of ways to explicitly rely on the any 'type' in TypeScript. Enough that it makes a search and delete not as trivial as you would think.
Too many people use the terms Microservice to refer to an independently deployed piece of code. This is not a good definition of a Microservice. Microservice architecture has nothing to do with deployment topology
Amazon ECS encourages us to adopt practices that conform with the twelve-factor app methodology
How to set up switch role permissions in AWS to easily move between accounts